On the other hand, if analysis detects a recoverable claim accurately, our experience of two decades in the industry tells us that the extent of the recovery possible is approximately 30% of the actual amount.
Clearly, the payment integrity function in healthcare is ripe for the application of AI such as Machine Learning, Deep Learning, Big Data Analytics, etc. Such predictive models will minimize revenue losses due to fraud and also improve provider experience, keeping NPS higher.
Smart and self-learning platform for payment integrity
Fortunately, there is a vast volume of data coming in from providers, members, labs, facilities, etc., which when coupled with readily available compute power for AI systems, is able to automatically and effectively recognize the patterns in these data sets, analyze outliers, and identify fraudulent claims. As the platform is fed with larger data sets, machine learning can be leveraged to improve existing analytical rules, flag risk factors, and assign risk scores to entities involved in making these predictive models even more efficient. Over a period of time, this system is autonomous, capable of self-learning and becoming smarter, delivering quantum improvements in the output.
At a minimum, once a claim is processed through such a self-learning platform, the AI platform should be able to detect member eligibility and benefits problems, contractual gaps, or pricing errors, etc., providing a targeted list of claims or providers that the Special Investigation Units (SIUs) can take up for further investigation. An AI system can also offer additional insight to analyze top providers at risk for errant billing, their relationship with a set of members and interaction with them. Such a system can also provide geospatial analysis of providers and their members. Such analysis is helpful in identifying members that are located far from providers and have high occurrences of visits for routine services. A dashboard offers a 360° view of a provider such as services offered, top areas served, performance over time around claim volume, amount paid, amount denied, anomalies detected in the past. Access to such insights serves as a powerful tool for the SIU during investigations.
Complementary heuristic intelligent analytics should discern many other unusual patterns of claims submission and processing, thus making it possible to distinguish between true positive versus false positives claim errors. If such systems can reduce the number of false positives or suspicious cases to be investigated by a mere 10-20%, it will afford the SIU the time to focus on a greater number of appropriate cases, which can result in higher recovery.
AI is applicable for the pre-payment cost avoidance model too
An emerging trend in the industry is the shift from a post-payment recovery to a pre-payment cost avoidance model. This model helps avoid 100% of false claims costs in contrast to recovering only around 30% in the post-payment model, provided the anomaly detection process is advanced and agile enough to detect the fraudulent claims before they are processed for payment.
The pre-payment review of claims also reduces the cost that goes into reworking, reprocessing and recovering claims. However, despite the advantages, payers are very cautious to adopt this model as there aren’t many good options. The primary reason for the caution lies in the fact that the model does not have a proven and well-demonstrated path to success. There is a limited window of time to examine a large volume of claims before they are processed and a delay in processing beyond statutory timelines could result in late payment penalties for the payers. The initial success we have seen shows that an AI based payment integrity function is uniquely positioned, in terms of agility and accuracy, to make the pre-payment cost avoidance model a reality soon.
Advancing into the future of payment integrity
While the pre-payment cost avoidance model has clear benefits, payers cannot make a 100% shift to this model due to its nascent nature. If they do, they will face varying challenges before the system becomes robust, quick and reliable. On the other hand, the post-pay model has known inefficiencies and needs modernization.
In the ideal scenario, a payer must seek to deploy both pre and post-payment models in tandem, maintaining a balance between the two without causing business disruption. Such an approach is possible when powered by an AI-based payment integrity platform and this is the inevitable future of the payment integrity function.